mcpbr
Verified Safeby greynewell
Overview
A benchmark runner for evaluating Model Context Protocol (MCP) servers by comparing LLM agent performance with and without MCP tools on software engineering tasks.
Installation
mcpbr run -c mcpbr.yamlEnvironment Variables
- ANTHROPIC_API_KEY
Security Notes
The server's core functionality involves executing user-defined MCP server commands and agent-generated code (patches, PoC exploits) within Docker containers. While Docker provides a layer of isolation, malicious configurations for the 'mcp_server' command or arguments could lead to undesirable actions within the container. The Claude Code CLI is run with '--dangerously-skip-permissions', though it is executed by a non-root user ('mcpbr') inside the container. The project's 'SECURITY.md' acknowledges these risks and advises users to only use trusted MCP servers and be aware of network access within containers.
Similar Servers
mcp-interviewer
A Python CLI tool to evaluate Model Context Protocol (MCP) servers for agentic use-cases, by inspecting capabilities, running functional tests, and providing LLM-as-a-judge evaluations.
1xn-vmcp
An open-source platform for composing, customizing, and extending multiple Model Context Protocol (MCP) servers into a single logical, virtual MCP server, enabling fine-grained context engineering for AI workflows and agents.
modular-mcp
A proxy server that efficiently manages and loads large tool collections from multiple Model Context Protocol (MCP) servers on-demand for LLMs, reducing context overhead.
mcp-framework
A Rust framework for building AI agents with built-in Model Context Protocol (MCP) support, multi-LLM integration, and a web-based inspector for debugging.